abhigyanpatwari/GitNexus

GitNexus is a zero-server code intelligence engine that transforms any GitHub repository or ZIP file into an interactive knowledge graph — entirely inside your browser. No backend required, no data leaves your machine. The tool solves a real problem every developer faces: understanding unfamiliar codebases. Instead of reading files one by one and building a mental model of how everything connects, GitNexus automatically maps every relationship — imports, function calls, class inheritance, module boundaries — into a visual, queryable graph. What makes GitNexus technically interesting is its seven-phase indexing pipeline: Structure, Parse, Imports, Calls, Heritage, Communities, and Processes. It precomputes clusters of related code, traces execution flows, and scores edge confidence at index time. When you query the graph, you get decision-ready context rather than raw file listings. GitNexus operates in two modes. The Web UI provides a quick way to drop in a repo and start chatting with it using the built-in Graph RAG agent. The CLI mode indexes your repository locally and exposes an MCP server, giving AI coding agents deep codebase awareness through the Model Context Protocol. The Graph RAG approach is particularly powerful for code review and refactoring decisions. Before changing a function, you can visualize its full blast radius — every caller, every dependency, every test that touches it. This transforms code exploration from a linear file-reading exercise into a spatial navigation experience. GitNexus runs entirely client-side using browser APIs, making it safe for proprietary codebases. The knowledge graph persists in the browser's IndexedDB, so subsequent visits to the same repo load instantly without re-indexing.